FutureHouse releases AI tools it claims can accelerate science

Future houseThe non -profit organization supported by Eric Schmidt aims to build a “scientist AI” in the next decade, has launched its first main product: platform and API with AI tools intended for supporting scientific work.

Many, many startups They are racing to develop AI research tools for scientific domain, some with huge amounts of VC financing behind them. Technology giants also look bull’s on science AI. Earlier this year, Google revealed “Ai co-scientific“Ai, the company said it could help scientists create hypotheses and experimental research plans.

Executive Directors of the Openi and Anthropic AI company have fixed In order for ALI tools to be able to accelerate scientific discovery, especially in drug. But many researchers do not consider AI to be particularly useful today conducting a scientific processlarge part because of its unreliability.

Futurehouse posted four tools on Thursday: Crow, Falcon, Owl and Phoenix. Crow can search for scientific literature and answer questions about it; Falcon can conduct deeper tests of literature, including scientific databases; OWL seeks previous work in a particular area; And Phoenix uses tools to plan chemistry experiments.

“Unlike others [AIs]Futurehouse have access to a huge corpus of high quality papers with open approach and specialized scientific tools, “” He writes Futurehouse in a blog post. “They [also] have a transparent explanation and use a multi -phase procedure to deeper to consider each source […] By authorizing these [AI]With together, scientists can greatly accelerate the pace of scientific discovery. “

But significantly, Futurehouse has yet to achieve a scientific breakthrough or make a new discovery with their ALI tools.

Part of the challenge in the development of “AI scientists”, anticipating an untold number of confusing factors. AI could come in handy in areas that need a broad research, such as narrowing a huge list of possibilities. But it is less clear whether AI is capable of solving the problem outside the box, which leads to Bonafide breakthrough.

The results of AI systems designed for science so far have been mostly insidious. Year 2023. Google said there were about 40 new materials synthesized With the help of one of his AIS, called Gnomes. More External analysis No material was found, in fact it was a net new one.

Ai -‘technical disadvantages and risks, such as his preference hallucinateAlso do scientists cautious that they support this for serious work. Even well -designed studies could eventually be extinguished by the wrong behavior of AI, which is struggling with the execution of highly precise work.

Indeed, Futurehouse admits that his alai tools – especially Phoenix – can go wrong.

“We are released [this] Now in the spirit of fast iteration, “the company writes in its blog post.” Please provide feedback while using it. “